A Wnt-producing niche drives proliferative potential and progression in lung adenocarcinoma

Abstract

The heterogeneity of cellular states in cancer has been linked to drug resistance, cancer progression and the presence of cancer cells with properties of normal tissue stem cells. Secreted Wnt signals maintain stem cells in various epithelial tissues, including in lung development and regeneration. Here we show that mouse and human lung adenocarcinomas display hierarchical features with two distinct subpopulations, one with high Wnt signalling activity and another forming a niche that provides the Wnt ligand. The Wnt responder cells showed increased tumour propagation ability, suggesting that these cells have features of normal tissue stem cells. Genetic perturbation of Wnt production or signalling suppressed tumour progression. Small-molecule inhibitors targeting essential posttranslational modification of Wnt reduced tumour growth and markedly decreased the proliferative potential of lung cancer cells, leading to improved survival of tumour-bearing mice. These results indicate that strategies for disrupting pathways that maintain stem-like and niche cell phenotypes can translate into effective anti-cancer therapies.

Extended Data Figure 1. Inhibiting Wnt synthesis with the Porcupine inhibitor LGK974 suppresses Wnt pathway activation by the R-spondin/Lgr5 axis in primary lung adenocarcinoma cultures

a, Percent of EdU+ (proliferating) cells in 3-dimensional (3D) cultures of Kras^G12D/+; Trp53Δ/Δ; Rosa26^tdTomato/+ (KPT) lung adenocarcinoma (LUAD) cells followed by Wnt pathway stimulation with Rspo1 (1 μg/ml), Wnt3a (100 ng/ml), or RW (1 μg/ml Rspo1 + 100 ng/ml Wnt3a), or Wnt pathway inhibition with LGK974 (100 nM) or DKK1 (500 ng/ml) for 7 days (starting at 7 days after plating). EdU labeling was performed 16h before analysis of proliferating cells by flow cytometry. N = 6 wells/condition. b, Percent of spheroids proliferating in low-density 3D cultures of primary mouse KPT LUAD cells 14 days after plating. N = 8 wells/condition. Representative data from 8 replicate experiments; TT5678 and TT5861 identify donor mice. c, Quantitative PCR of Porcn transcripts in sublines of a KP LUAD cell line expressing shRNAs targeting Porcn (shPorcn) or control shRNA (shLuc). Quantification of 3D tumour spheroids containing EdU+ cells of KP LUAD cells expressing the indicated shRNAs in response to 100 ng/ml Wnt3a or control at 6 days after plating. Representative data of n = 3 technical replicates, the experiment was performed with three independent cell lines. e, Quantitative real-time PCR (qRT-PCR) analysis of Axin2 and Lgr5 transcripts in 3D cultures of primary Kras^G12D/+; Trp53Δ/Δ (KP) LUAD cells following Wnt pathway stimulation with Rspo1 (1 μg/ml), Wnt3a (100 ng/ml), or RW (1 μg/ml Rspo1 + 100 ng/ml Wnt3a), or Wnt pathway inhibition with LGK974 (100 nM) or DKK1 (1 μg/ml) for 6 days (starting at 10 days after plating). Representative data of n = 3 technical replicates, the experiment was carried out three times, each time with cells isolated from different tumor-bearing mouse. f, Quantification of tumour spheroids containing EdU+ cells per 100 primary KPT LUAD cells 14 days after plating. RW (1 μg/ml Rspo1 + 100 ng/ml Wnt3a); LGK974 (100 nM). N = 8 wells/condition. g, Recipient mouse lungs 8 weeks following orthotopic transplantation of 30,000 primary tdTomato+ (red) primary mouse KP LUAD cells. Arrowheads indicate tdTomato+ tumours. Recipient mice were treated with 10 mg/kg/d LGK974 or vehicle for 8 weeks, starting from the day of transplantation. Scale bar: 2 mm. The experiments was performed three times, each time with cells isolated from different (donor) tumor-bearing mouse. h, Wnt pathway activity measured by detection of firefly luciferase driven by a Wnt-sensitive 7TCF promoter (TOPFLASH assay) in murine KP LUAD cells stimulated for 24h with the indicated growth factors. N = 3 technical replicates/condition. Representative data from experiments that were performed with four different cell lines. i, j, k, Quantitative PCR analysis of Rspo2 (i), Rspo3 (j) or Lgr5 (k) transcripts in two independent murine KP LUAD cell lines (TT5734 and TT5764) expressing Synergistic Activation Mediator (SAM) components driving expression of the indicated genes. l, Flow cytometry analysis of GFP fluorescence in a KP LUAD cell line harboring the Lgr5^GFP-CreER reporter expressing vector control (top panel) or an sgRNA targeting the transcription start site of Lgr5 (bottom panel). m, n, qRT-PCR analysis of the Wnt target genes Axin2 (m) and Lgr5 (n) in 3D cultures of sublines of a KP LUAD cell line (TT5764) expressing the CRISPR-activator system driving expression of Rspo2 (Rspo2-a), Rspo3 (Rspo3-a) or Lgr5 (Lgr5-a). * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001. Two-way ANOVA: (a), (b), (d), (h), (m), (n); Student’s two-sided t-test: (c), (e), (f), (i), (j), (k); error bars: +SD.

Extended Data Figure 2. Lgr4 and Lgr5 are R-spondin receptors in lung adenocarcinoma

a, Quantitative real-time PCR (qRT-PCR) analysis of Lgr4, Lgr5, and Lgr6 transcripts in sublines of a Kras^G12D/+; Trp53Δ/Δ (KP) LUAD cell line stably expressing shLgr4, shLgr5, shLgr6, or shLuciferase (shLuc) control. Note minimal effects of the indicated shRNAs on other Lgr5 family members. N = 3 technical replicates/condition; representative data from 3 experiments carried out in different cell lines. b, c, Formation of 3D tumour spheroids of sublines of a KP LUAD cell line expressing the indicated shRNAs in response to 1 μg/ml Rspo1 (b, c) or 100 ng/ml Wnt3a (d, e) 6 days after plating. Scale bars: 500 μm. N = 8 wells/condition, representative data from 3 experiments carried out in different cell lines. f, g, No difference in cell morphology (f) or growth rate (g) in sublines of a KP LUAD cell line expressing shLgr4, shLgr5, shLgr6, or control shLuciferase (shLuc) over 6 days in 2-dimensional cell culture. Scale bar: 200 μm. * P < 0.05; ** P < 0.01; *** P < 0.001. Student’s two-sided t-test: (a); two-way ANOVA: (c), (e), (g); error bars: +SD. All experiments in this figure were performed three times, each time with a different KP LUAD cell line.

Extended Data Figure 3. Wnt ligands produced predominantly by cancer cells drive activation of the Wnt signaling pathway in lung adenocarcinoma

a, Schematic representation of the lentivirus vector used to transduce a Kras^G12D/+; Trp53Δ/Δ; Rosa26^tdTomato/+ (KPT) LUAD cell line, followed by puromycin selection. b, tdTomato and 7TCF::Luciferase signals at baseline (0h) and 48h following treatment with 10 mg/kg/d LGK974 or vehicle. Red arrows indicate two subcutaneous tumours with reduced 7TCF-dependent bioluminescence in response to 48h of LGK974 treatment. c, Suppression of 7TCF-driven bioluminescence by LGK974 relative to tdTomato signal in mice harboring subcutaneous transplants of the KPT LUAD cell line stably expressing the 7TCF::Luciferase-PGK-Puro lentivirus. N = 6 tumours, 3 mice per group. Representative data from two independent experiments. Student’s two-sided t-test; error bars: ±SD. d, Haematoxylin-eosin (HE) or β-catenin staining in KP adenomas or in adenocarcinomas. Scale bars: 100 μm. e, Immunofluorescence for β-catenin (red) and Porcupine (green) in a Kras^G12D/+; Trp53Δ/Δ (KP) lung adenocarcinoma (LUAD). Scale bar: 100 μm. f, In situ hybridization (ISH) for Axin2 mRNA (purple, arrowheads) in KP lung adenomas or adenocarcinomas. Scale bar: 50 μm. g, Immunofluorescence for Porcupine (green, white arrowheads) and ISH for Axin2 (red, yellow arrowheads) in KP LUAD. Scale bar: 10 μm. h, Immunofluorescence for Porcupine (green) in an autochthonous grade 3 KPT adenocarcinoma. Arrowheads indicate peritumoural Porcupine+ cells, which are tdTomato negative (i.e. not cancer cells). Scale bar: 100 μm. i, CD11b (green) and Porcupine (red) immunofluorescence in a peritumoural region. White line delineates tumour (T). Scale bar: 100 μm. j, Immunohistochemistry for β-catenin or Porcupine in human LUAD. Arrowheads indicate cells with nuclear β-catenin. Scale bar. 100 μm. 65 human LUAD tumors in two tissue microarrays were analysed. k, Comparison of PORCN gene expression in tumours versus normal tissue in The Cancer Genome Atlas (TCGA) lung adenocarcinoma cohort: Empirical cumulative density function (CDF) plots of standardized gene expression values are shown. A right-shift indicates relatively higher expression, with P-values indicated to assess statistical significance (KS-test).

Extended Data Figure 4. Expression of Porcupine in lung adenocarcinoma, the normal lung and in stem cell niches of the intestine and liver

a, In situ hybridization (ISH) for Porcn in a grade 3 Kras^G12D/+; Trp53Δ/Δ (KP) lung adenocarcinoma. b, c, Immunohistochemistry for Porcupine (brown) (b), or Porcn in situ hybridization (ISH) in the normal lung (c). Arrowheads indicate Porcupine expression in bronchioles. d, Immunohistochemistry for β-catenin (brown) and Porcupine (purple) in the small intestine. Note Porcupine expression in intestinal crypts (black arrowheads), the location of Lgr5+ intestinal stem cells, as well as in transit-amplifying cells (blue arrowheads) and stromal cells (dark green). e, Immunostaining for Porcupine (brown) and glutamine synthetase (GS, purple) expression localizes to areas around the central vein of the liver (e), coinciding with the location of Lgr5+/Axin2+ liver stem cells^,. Scale bars: 100 μm.

Extended Data Figure 5. Analysis of the Porcn locus following CRISPR/Cas9-mediated genome editing in vivo

a, Haematoxylin-eosin staining of KP LUAD-bearing lungs generated with pSECC-sgTom or pSECC-sgPorcn and quantification of the proportion of adenomas vs. adenocarcinomas at 12 weeks following tumour initiation. Scale bar: 2 mm. Student’s two-sided t-test; error bars: +SD. N = 5 mice per group. b, Massively parallel sequencing analysis of allelic fractions of the Porcn locus in lung lobes containing microscopic tumours (“sgPorcn.2 L”) or microdissected macroscopic tumours (“sgPorcn.2 T”) induced in Kras^LSL-G12D/+; Trp53^flox/flox (KP) mice using pSECC-sgPorcn.2, or in lung lobes or macroscopic tumours induced in KP mice using pSECC-sgTom.2 (“sgTom.2 T/L”). WT: wild-type read; FS: frameshift mutation; NFS: non-frameshift mutation; Low freq: low-frequency mutation event. Note predominantly wild-type or non-frameshifting reads in microdissected tumours, whereas mutations in tumours containing microscopic tumours have introduced frameshifts. The large contribution of wild-type reads in “sgPorcn.2 L” samples is due to domination of the normal stroma in whole-lobe samples, whereas wild-type reads in sgPorcn.2 T indicate cancer cells where genome editing did not function, as in whole tumour samples tumour cells are expected to contribute at least 50% (ref. 49). c, Qualitative analysis of mutations introduced by sgPorcn.2 in vivo. INS: insertion; DEL: deletion; bp: base pair (indicates size of insertion/deletion). Ratio indicates frequency of event across 15 samples analysed.

Extended Data Figure 6. Lgr5 and Lgr4 are expressed in lung adenocarcinoma, and Lgr5 marks cells with increased tumour-forming ability

a, Quantitative PCR analysis of Lgr5 gene expression in Kras^G12D/+; Trp53Δ/Δ (KP) LUAD tumours microdissected at 9 weeks (adenomas) or 20 weeks (adenocarcinomas) post-initiation with adenoviral Cre. N = 6 tumors/group. b, In situ hybridization (ISH) for Lgr5 or Lgr4 mRNA (purple) in grade 3 KP LUAD adenocarcinomas 12 weeks following tumour induction with AdCre. Scale bars: 100 μm. c, d, FACS sorting (d) of Lgr5+ (GFP+) cells in cultured KP LUAD cell line containing the Lgr5^GFP-CreER/+ reporter allele (KP; Lgr5^GFP-CreER/+), followed by quantitative real-time PCR analysis (c) of Lgr4 expression in Lgr5+ cells in two independent cell lines (TT1937 and TT6280). This experiment was performed once. e, Fluorescence-assisted cell sorting (FACS) of GFP+ and GFP- cells isolated from KPT; Lgr5^GFP-CreER/+ primary LUAD 14 weeks following tumour initiation with intratracheally administered AdCre. The FACS plot is gated on tdTomato+/CD11b−/CD31−/CD45−/TER119− cells. Note bleeding of the tdTomato signal to the GFP channel in the panel on the right. Gates were drawn as shown to increase cell yield at the cost of purity to enrich for Lgr5+ cells. Such FACS sorting was performed on 21 KPT; Lgr5^GFP-CreER/+ mice. f, Recipient mouse lungs 12 weeks following orthotopic transplantation of 15,000 primary Lgr5+/tdTomato+ or Lgr5−/tdTomato+ primary mouse LUAD cells. Arrowheads indicate tdTomato+ tumours (red). Scale bar: 2 mm. Representative data from three replicate experiments. g, Quantification of tumours per 1000 cells in recipient mouse lungs 12 weeks following orthotopic transplantation of 15,000 primary Lgr5+/tdTomato+ or Lgr5−/tdTomato+ primary mouse LUAD cells. N = 3 recipient mice per group; representative data from three replicate experiments. *P < 0.05. h, Number of membrane-associated GFP+ (mG+) clones following lineage-tracing in established subcutaneous KP; Lgr5^CreER/+; Rosa26^LSL-mTmG/+ LUAD primary transplants (see Figure 3c). N = 9 tumors/time point. Student’s two-sided t-test: (a), (c), (g), (h). *P < 0.05, n.s. = not significant: error bars: +SD.

Extended Data Figure 7. Phenotypical plasticity of Lgr5+ cells that reside in Porcupine+ niches in lung, pancreatic and colon tumours

a, Immunofluorescence for GFP (green) and EpCAM (red) in a subcutaneous transplant established from a single-cell clone of Kras^G12D/+; Trp53Δ/Δ; Lgr5^CreER/+ cell line. Scale bar: 1 mm. b, Immunofluorescence for GFP (green) and Porcupine (red) in a subcutaneous transplant established from a single-cell clone of Kras^G12D/+; Trp53Δ/Δ; Lgr5^CreER/+ cell line. (a, b) are representative data from four replicate experiments, each with a different Kras^G12D/+; Trp53Δ/Δ; Lgr5^CreER/+ cell line. c, ISH for Lgr5 mRNA (purple) in KP; Pdx1::Cre pancreatic ductal adenocarcinoma (PDAC). Scale bars: 100 μm (top) and 10 μm (bottom). Representative data from 3 PDAC tumors analysed. d, Immunofluorescence staining for GFP (green) in a tdTomato+ (red) autochthonous KP; Lgr5^GFP-CreER/+; Rosa26^tdTomato/+; Pdx1::Cre PDAC. Scale bars: 100 μm (top) and 10 μm (bottom). e, Quantification of primary spheroids containing EdU+ cells per 100 Lgr5+/tdTomato+ or Lgr5−/tdTomato+ primary mouse PDAC cells plated. N = 4 wells/group. *P < 0.05, Student’s two-sided t-test; error bars: +SD. f, Immunofluorescence staining for GFP (green) and Porcupine (red) in autochthonous KP; Lgr5^GFP-CreER/+; Pdx1::Cre PDAC. Note juxtaposition of Lgr5+ and Porcupine+ cells in the tumours. Scale bars: 100 μm (top) and 10 μm (bottom). (d, f) Representative data from 6 KP; Lgr5^GFP-CreER/+; Pdx1::Cre PDAC tumors analysed. g, Immunofluorescence staining for GFP (green) and Porcupine in an autochthonous ApcΔ/Δ; Lgr5^GFP-CreER/+intestinal adenoma. Again, note juxtaposition of Lgr5+ and Porcupine+ cells in the tumours. Scale bar: 100 μm. N = 3 tumor samples. h, Immunohistochemistry for Porcupine (brown) in human colorectal adenocarcinoma. Scale bars: 100 μm (top) and 10 μm (bottom). 5 human colorectal adenocarcinoma samples were analysed.

Extended Data Figure 8. Wnt pathway activation correlates with poor survival in human lung adenocarcinoma, pancreatic ductal adenocarcinoma and mesothelioma, but not in human squamous cell lung cancer; analysis of the Lgr4 and Lgr5 loci following CRISPR/Cas9-mediated genome editing in vivo

a, Kaplan-Meier survival curve comparing the 20% strongest (red, n = 91) and weakest (blue, n = 92) Willert Wnt signature correlated patients from the TCGA Lung Adenocarcinoma cohort. b, Empirical cumulative density function (CDF) plots of standardized gene expression values showing a correlation between the Wnt pathway activation gene expression signature correlation score and histological grade of primary tumours. A right-shift indicates relatively higher expression, with P-values indicated to assess statistical significance (KS-test). c-e, Kaplan-Meier survival curvex comparing the 20% strongest (red) and weakest (blue) Willert Wnt signature correlated patients from the TCGA Squamous Cell Lung Cancer (most-correlated/red n = 100, least-correlated/blue n = 100) (c), Pancreatic Ductal Adenocarcinoma (most-correlated/red n=34, least-correlated/blue n=34) (d), and Mesothelioma (most-correlated/red n=17, least-correlated/blue n=18) (e) cohorts. f, Massively parallel sequencing analysis of allelic fractions of the Lgr4 and Lgr5 loci in lung lobes containing microscopic tumours (“Lobe”) or microdissected macroscopic tumours (“Tumour”) induced in Kras^LSL-G12D/+; Trp53^flox/flox; Rosa26^{LSL-Cas9+2a+eGFP/+} mice using hU6::sgLgr4-sU6::sgLgr5-EFS::Cre (pU2SEC) or hU6::sgTom-EFS::Cre (pUSEC) lentiviral vectors. WT: wild-type read; FS: frameshift mutation; NFS: non-frameshift mutation. Note predominantly wild-type or non-frameshifting reads in microdissected tumours, whereas mutations in tumours containing microscopic tumours have introduced frameshifts. The large contribution of wild-type reads in “Lobe” samples is due to domination of the normal stroma in whole-lobe samples, whereas wild-type reads in Lgr4/Lgr5 co-targeted tumours indicate cancer cells where genome editing did not function, as in whole tumour samples tumour cells are expected to contribute at least 50% (ref. 49). g, h, Qualitative analysis of mutations introduced by sgLgr4 or sgLgr5 in vivo. INS: insertion; DEL: deletion; bp: base pair (indicates size of insertion/deletion). Ratio indicates frequency of event across all samples analysed. P-values are indicated in the figure (log-rank test).

Extended Data Figure 9. Characterization of the niche for stem-like cells in lung adenocarcinoma

a, Quantitative PCR analysis of Rspo gene expression in 16 KP LUAD tumours, normalized to Actb expression. Tumours were harvested at 16 weeks post-initiation with adenoviral Cre. In situ hybridization (ISH) for Rspo1 mRNA (purple, arrowheads) in a KP LUAD tumour. Note Rspo1 transcripts in endothelial cells. b, qPCR for Rspo1 and Rspo3 in tdTomato+ tumour cells (Tumour), CD31+ endothelial cells and the rest of the cells (Stroma) in microdissected Kras^G12D/+; Trp53Δ/Δ; Rosa26^tdTomato/+ (KPT) LUAD tumours following sorting. The expression of Pecam1 (that encodes CD31) was found to be >400-fold enriched in the CD31+ fraction compared to the stroma (not shown). N = 3 mice, representative from two replicate experiments. c, Heatmap showing relative expression levels of Porcn and the 19 murine Wnt genes based on quantitative PCR analysis in sorted tdTomato+ KP LUAD cells (T) vs. tdTomato- stromal cells (S) in microdissected tumours harvested at 20 weeks post tumour initiation (a time point when most tumours are adenocarcinomas). d, Volcano plot of qPCR array gene expression analysis showing statistically significant differentially expressed genes (in red, Fzd receptors are circled). X-axis is the log2 fold-change (Tumour/Stroma) and Y-axis is the –log10 P-value of the differential enrichment (2-sided t-test). e, Quantitative PCR analysis of Wnt5a, Wnt7a and Wnt7b gene expression in KP tumours microdissected at 9 weeks (adenomas) or 20 weeks (adenocarcinomas) post-initiation with adenoviral Cre. N = 6 mice, representative from two replicate experiments. f, Comparison of WNT gene expression in tumours versus normal tissue in The Cancer Genome Atlas (TCGA) lung adenocarcinoma cohort: Empirical cumulative density function (CDF) plots of standardized gene expression values for WNT5A, and WNT7B are shown. A right-shift indicates relatively higher expression, with P-values indicated to assess statistical significance (KS-test). g, Heatmap showing relative expression levels of Lrp5, Lrp6 and 9 murine Fzd genes based on quantitative PCR analysis in sorted tdTomato+ KP LUAD cells (T) vs. tdTomato- stromal cells (S) in microdissected tumours harvested at 20 weeks post tumour initiation (a time point when most tumours are adenocarcinomas). h, Quantitative PCR analysis of 8 Fzd receptors in KP tumours microdissected at 9 weeks (adenomas) or 20 weeks (adenocarcinomas) post-initiation with adenoviral Cre. N = 3 mice. Student’s two-sided t-test: (b), (e), (h); *P < 0.05; error bars: +SD.

Extended Data Figure 10. Porcupine inhibition suppresses Wnt pathway activity, progression and proliferative potential in autochthonous murine KP lung adenocarcinomas

a, Quantitative PCR (qPCR) analysis of Axin2 and Lgr5 transcripts in Kras^G12D/+; Trp53Δ/Δ (KP) LUAD tumours 2 weeks following treatment with 10 mg/kg/d LGK974 or vehicle. Treatment was started at 11 weeks post-tumour initiation. N = 6 tumors/group. b, Quantification of μCT data showing change in tumour volume compared to baseline (obtained at 76 days post tumor initiation, dashed line) after 4 weeks of 10 mg/kg/d LGK974 or vehicle control. c, Recipient mouse lungs 4 weeks following orthotopic GEMM-DA of 50,000 primary tdTomato+ (red) primary mouse LUAD cells. Arrowheads indicate tdTomato+ tumours. Donor mice bearing autochthonous Kras^G12D/+; Trp53Δ/Δ; Rosa26^tdTomato/+ (KPT) LUAD tumours were treated for 2 weeks with LGK974 or vehicle (black box, starting at 84 days post-tumour induction). The recipient mice were treated with LGK974 or vehicle (white box) for 4 weeks. Scale bar: 2 mm. d, tdTomato+ tumours in sections from lungs in (c) containing EdU+ cells (white arrowheads) or not containing EdU+ cells (yellow arrowheads). Scale bars: 500 μm (top) and 100 μm (bottom). e, Quantification of EdU+ (black) or EdU- (gray) tumours per section through the lungs depicted in (c, d). N = 5 (Vehicle-Vehicle), epresentative data from three replicate experiments. *P < 0.05; Student’s two-sided t-test: (a, b); two-way ANOVA (e); error bars: ±SD.

Figure 1. Ligand-dependent Wnt signaling sustains proliferative potential in lung adenocarcinoma

a, 3D cultures of sorted tdTomato+ (red) primary mouse Kras^G12D/+; Trp53Δ/Δ; Rosa26^tdTomato/+ (KPT) LUAD cells 14 days after plating. Proliferating (EdU+) cells (green, arrowheads). Scale bar: 100 μm. b, Quantification of tumour spheroids containing EdU+ cells from two mice (TT5678 and TT5861). N = 8 wells/condition. c, Quantification of KPT LUAD primary transplant tumours in recipient mouse lungs treated with LGK974 or vehicle for 8 weeks. d, Quantification of tumour spheroids containing EdU+ cells 10 days after plating. Rspo2-a, Rspo3-a and Lgr5-a refer to sublines expressing CRISPR-activator (SAM) components driving expression of the indicated gene. N = 8 wells/condition. e, Wnt pathway activity measured by TOPFLASH assay in KP LUAD sublines stably expressing shRNAs targeting Lgr4, Lgr5, Lgr4+Lgr5 or a Vector control. N = 3 technical replicates/condition, experiment was repeated 4 times. *P < 0.05; **P < 0.01; ***P < 0.001 compared to control, except in (d) where comparison in the LGK974 group is to the same CRISPR-activator line, and in (e) where comparison is to Rspo1 stimulation. Two-way ANOVA: (b), (d–e); Student’s t-test: (c); error bars: +SD.

Figure 2. Porcupine+ cancer cells form a niche that drives Wnt signaling in lung adenocarcinoma

a, Bioluminescence in lungs of Kras^LSL-G12D/+; Trp53^flox/flox (KP) or Kras^LSL-G12D/+; Apc^flox/flox mice 20 weeks following intratracheal infection with 7TCF::Luciferase-PGK::Cre lentivirus (see schematic). Scale bar: 5 mm. b, GFP (green) and EpCAM (red) staining in KP LUAD tumours 20 weeks following infection with 7TCF::GFP-PGK::Cre lentivirus (see schematic). Arrowheads indicate GFP+ cells with 7TCF promoter activity. Scale bars: 100 μm (left) and 10 μm (right). c, Haematoxylin-eosin (HE), β-catenin or Porcupine staining in KP adenomas or in adenocarcinomas. Note nuclear localization of β-catenin in adenocarcinomas (arrowheads). Scale bars: 100 μm. d, Percent of tumours containing subpopulations of cells with Porcupine expression or nuclear β-catenin per histological grade, or in metastases (M). N = 19 grade 1, 47 grade 2, 31 grade 3 and 11 grade 4 tumours, and 6 lymph node or thoracic wall metastases. e, Quantitative PCR analysis of Porcn gene expression in KP tumours at 9 weeks (adenomas) or 20 weeks (adenocarcinomas) post-initiation. ***P < 0.001, n = 16. f, Percent of human lung adenocarcinomas with absent, heterogenous or uniform Porcupine expression or nuclear β-catenin. N = 65. g, Quantification of tumour burden in KP mice 12 or 20 weeks following infection with 25,000 TU of pSECC-sgTom or pSECC-sgPorcn. h, Haematoxylin-eosin staining of KP LUAD-bearing lungs generated with pSECC-sgTom or pSECC-sgPorcn and quantification of the proportion of adenomas vs. adenocarcinomas at 20 weeks following tumour initiation. Scale bar: 2 mm. *P < 0.05. Student’s two-sided t-test: (e, g–h); error bars: +SD.

Figure 3. Lgr5+ lung adenocarcinoma cells display persistent proliferative potential

a, Immunostaining for GFP (green), Porcupine (red), and EpCAM (blue) in a subcutaneous transplant of primary KP; Lgr5^GFP-CreER/+ LUAD cells 3 weeks following transplantation. Scale bars: 100 μm (left) and 10 μm (right). Quantification of the percent of Lgr5+ cells that are adjacent to Porcupine+ or Porcupine- cells, or that are positive for both Lgr5 and Porcupine in the transplants. N = 6 tumors. b, 3D culture of Lgr5+ and Lgr5- KP LUAD cells. Scale bar: 100 μm. Quantification of primary spheroids containing EdU+ cells (arrowheads). N = 8 wells/group. c, Lineage-tracing of of KP; Lgr5^CreER/+; Rosa26^LSL-mTmG/+ LUAD cells in established subcutaneous primary transplants. Note migration of individual mG positive cells (yellow arrowheads) away from clones derived from Lgr5+ cells, and Porcupine+ progeny arising from Lgr5+ cells (white arrowheads). N = 9 tumours/group. Quantification of average clone size at 2 and 14 days post-tamoxifen administration. d, Quantification of tumour burden and proportion of adenomas/adenocarcinomas, and haematoxylin-eosin staining in KP; Rosa26^{LSL-Cas9-2a-GFP/+} mice infected with the indicated lentiviral vectors. hU6: human U6 promoter; sU6: synthetic U6 promoter; EFS: minimal E1α promoter. N = 6 (sgTom), 7 (sgLgr + sgLgr5). Scale bar: 2 mm. e, mRNA ISH for the indicated transcripts (purple) in consecutive sections of a similar region of a KP lung adenocarcinoma. Scale bar: 100 μm. Student’s two-sided t-test: (a–d); *P < 0.05, **P < 0.01; error bars: ±SD.

Figure 4. Porcupine inhibition improves survival by suppressing proliferative potential in mice harboring lung adenocarcinoma

a, Fold change in bioluminescence signal in autochthonous KP LUAD tumours harboring a Rosa26^Luciferase/+ allele in mice treated with LGK974 or vehicle. N = 3. b, Survival of mice harboring autochthonous KP LUAD tumours treated with LGK974 or vehicle starting at 77 days following tumour initiation. P = 0.0008; n = 5 (LGK974), 8 (Vehicle). c, Lung μCT images of mice treated with vehicle or LGK974 at 77 days after tumour initiation and after 40 days on therapy (Day 117). H: heart, T: tumour. d, Quantification of proliferating (Ki67+) cells in autochthonous KP LUAD tumours 2 weeks following treatment with LGK974 or vehicle. ****P < 0.0001; n = 80 vehicle tumours, n = 59 LGK974 tumours. e, Quantification of the number of tdTomato+ surface tumours in recipient mice. Student’s two-sided t-test: (a), (b), (d); two-way ANOVA: (e); error bars: ±SD. f, The outcome of Wnt inhibition in LUAD: Porcupine+ niche cells provide Wnt to Lgr5+ cells with robust proliferative potential, which can be suppressed by Wnt inhibitors. g, The niche for Wnt responder cells in LUAD: Lgr5+ cells (green) reside next to Porcupine+ cells (red). Wnt5a, Wnt7a and Wnt7b, provided by Porcupine+ niche cells (red), bind Frizzled on Wnt responder cells. Rspo1 and Rspo3, which bind Lgr4 and Lgr5, reinforce Wnt signaling by inhibiting Rnf43 and Znrf3 ubiquitin ligases that degrade Frizzled. Wnt is palmitoylated (serrated line) by Porcupine, critical for Wnt secretion and binding to Frizzled.